Prediction, Knowledge, And Explainability: Examining The Use of General Value Functions in Machine Knowledge